Research

EvoGenUI-Bench: Evaluating LLMs as Multi-Turn Generative UI Assistants

Yue Peng, Lanke Xia, Zihan Wang, Jiahao Ye, Ke Ning, Hongyi Wen

EMNLP 2026 Main Conference

Tests whether multi-turn generative UI assistants preserve executable requirements, tool grounding, and text-interface alignment across revisions.

TFP: Temporally Conditioned Memory-Fusion Policies for Visuomotor Learning

Yushen Liang*, Yue Peng*, Baosheng Jin*, Tianluo Zhang, Xinyu Zhang, Shuyi Zhou, Zhuoran Chen, Xinqi Liu, Shenji Wan

* Equal contribution

SemRob Workshop, Robotics: Science and Systems 2026

An event-sensitive task belief conditions flow-matching action generation for long-horizon manipulation.

FDB&FragLinker: A Large Fragment Database for Rapid Ligand Optimization Within Protein-Ligand Complex

Lei Zheng, Qisheng Zhou, Tianxiang Fu, Yue Peng, Yizhe Dai, Rong Chen, Zhaoxi Sun, John Z. H. Zhang

Journal of Molecular Biology, 438(18), 169741

Physical Time as a Missing Dimension of Robot Policy Memory Dynamics

Yushen Liang, Yue Peng, Baosheng Jin, Tianluo Zhang, Xinyu Zhang, Shuyi Zhou, Zhuoran Chen, Xinqi Liu, Shenji Wan

Under review at ICLR 2027

Temporal Flow Policy retains past events and evolves its memory over measured time intervals to guide robot actions in tasks requiring timing and memory.

G³VLA: Geometric Inductive Bias for Vision-Language-Action Models

Yue Peng, Yongzhe Zhao, Artur Habuda, Khuyen Pham, Yanheng Zhu, Tran Nguyen Le, Fares Abu-Dakka, Li Guo

Under review at IEEE Robotics and Automation Letters (RA-L)

A camera-aware module that adds ray geometry, projective positional encoding, and cross-view fusion to pretrained VLA visual tokens.

NovelAPIBench: Diagnosing How A Code LLM Learns to Use Novel APIs

Jinnuo Liu, Yue Peng*, Jinhan Niu*, Hongyi Wen

* Equal contribution

Under review at ICLR 2027

NovelAPIBench generates executable tasks for unseen APIs and diagnoses failures across six categories of API knowledge.